Kepler Nvidia GeForce GTX 780

[Tim]

Kyle expects the 680 to be faster at this point in time.

If and when we finally see the GK110 its likely going to be a monster performer. AMD may really have set the standard to low this time.

Okay, didn't he say a few posts back he knew how fast it is?

[GoldenTiger]

Kyle claims: http://hardforum.com/showpost.php?p=...9&postcount=89

You will start to see benchmark leaks and such probably in the next couple of days. I think most of those will be at less than 2560 resolution. And let's face it, you don't need a flagship card for resolution that small. Honestly, like always, we are going to have to put it through [H] large resolution gaming paces before we make a call on where it is compared to 7970. I think the 680 will be faster but by how much, we don't know yet, and I have seen NO MSRPs confirmed.

[UrbanSmooth]

The rising tension of this thread is really starting to build...

[Dimitriman]

Show me the card!!!
I going to the great country United States of America next month, second best country after Kazakhstan... and I going to buy brand new PeeCees! Very niiice!
Choosing between 7870 OC or 680.. please bring it out soon!

[DilTech]

How can you believe this
@Damien I totally agree with you

I have a feeling it's a mis-quote and they meant 200w for the video card. That part may be believable.

[boxleitnerb]

This dynamic clocking rumor is very interesting. I hope it's sophisticated. I could imagine the TDP budget being shifted around on the GPU as needed. Highly utilized units are clocked higher while waiting/idle units are clocked lower. This would increase power consumption on average but would stay within the TDP. I hope one can deactivate it for older games though. Or when you are CPU bound or have vsync enabled that it doesn't kick in when the chip doesn't run at full capacity.

[G.Foyle]

I wish good luck for reviewers for bench the card, know at what speed is running the card, and what happend during this time.. is the card at 750-780-800-900? what will be the result if you bench the card after 1 hour of play ? will it be the same after the card got high heat? Is it OC by App for gain some fps in benchmark, is the result will be the same when someone play 1hour of BF3? Does it detect games benchmark and adapt the OC profile, but is it working the same when gaming ?

It's been the same thing with benchmarking CPUs with turbo mode. You run the bench 10 times time after time and you get 10 different results... numbers in reviews are more like close approximations, that's why a score difference of 1-3% between competing products is called a draw, not a win.

[Handrox]

relative to 7970 for the advantages of the new stuff is as follows: low-voltage, high frequency (broken GHz), low power consumption low-noise, high-performance (the DX11), AA, and obtained the support of the business of the game

PHK Expreview -> http://bbs.expreview.com/thread-49806-1-1.html

[labs23]

GTX6xx pics, teh card, another PCB shot, and those power-connectors.

Source:http://we.pcinlife.com/thread-1849001-1-1.html

[LordEC911]

This dynamic clocking rumor is very interesting. I hope it's sophisticated. I could imagine the TDP budget being shifted around on the GPU as needed. Highly utilized units are clocked higher while waiting/idle units are clocked lower. This would increase power consumption on average but would stay within the TDP. I hope one can deactivate it for older games though. Or when you are CPU bound or have vsync enabled that it doesn't kick in when the chip doesn't run at full capacity.

Do you realize what sort of tranny budget that would add?
Nvidia didn't even want to go simple for Fermi, they went software to try to keep the cards under control and we all know how well that worked out...

Not saying it is impossible but I would be very surprised to see them jump into that all at once rather than gradually implementing the feature over a generation or two.
I see them implementing into the hardware to control TDP, such as AMD's Powertune, but anything more seems unlikely at this point especially something relatively complex.

[boxleitnerb]

Several sources speak of decoupled shader and core clocks. This is already different from Powertune. It seems, both clock domains are dynamic (705-950MHz for the core and up to 1411 MHz for the ALUs).

[N19h7m4r3]

As long as those connectors don't make the card thicker thana standard 2 slot design I don't care.

If they do, well others and myself might have trouble getting it into a M-ITX system.

[LordEC911]

Several sources speak of decoupled shader and core clocks. This is already different from Powertune. It seems, both clock domains are dynamic (705-950MHz for the core and up to 1411 MHz for the ALUs).

It seems I have missed a few articles over the last 24hours. Hard trying to stay up to date with this silly season.

[jam2k]

[GoldenTiger]

ROFL, well done.

[jam2k]

http://www.rage3d.com/board/showpost...postcount=3873

[G.Foyle]

Several sources speak of decoupled shader and core clocks. This is already different from Powertune. It seems, both clock domains are dynamic (705-950MHz for the core and up to 1411 MHz for the ALUs).

This is nonsense :P So the shaders would work at less than 2:1 clock ratio? That might have been possible two GPU generations ago.

[boxleitnerb]

Why nonsense? Please elaborate and remember that we're talking about 1536 of those bad boys here, not just 512 or so.

[G.Foyle]

Ask yourself:
1) what kind of clocks can Nvidia expect from TSMC 28 nm?
2) what kind of transistor budget they have?
3) what was the shader clock in every generation of GPUs that had hot clocks?

And finally:
4) is it therefore possible to have 1536 SPs at 1500+ MHz and still fit into thermal and power constraints?

[El Mano]

MWA HA HA A HA
Awesome!

[trinibwoy]

Ask yourself:
1) what kind of clocks can Nvidia expect from TSMC 28 nm?
2) what kind of transistor budget they have?
3) what was the shader clock in every generation of GPUs that had hot clocks?

And finally:
4) is it therefore possible to have 1536 SPs at 1500+ MHz and still fit into thermal and power constraints?

When you saw G70 did you think G80 was possible?

[SKYMTL]

Kyle claims: http://hardforum.com/showpost.php?p=...9&postcount=89

I beg to differ.

It is nearly impossible to play games like BF3, Skyrim (plus official texture pack), Batman AC and several other AAA titles with image quality maximized WITHOUT a flagship card at 2560.

[GoldenTiger]

I beg to differ.

It is nearly impossible to play games like BF3, Skyrim (plus official texture pack), Batman AC and several other AAA titles with image quality maximized WITHOUT a flagship card at 2560.

Agreed, though I think he was referring to 1920x1080... at least I hope he was haha.

[DilTech]

I beg to differ.

It is nearly impossible to play games like BF3, Skyrim (plus official texture pack), Batman AC and several other AAA titles with image quality maximized WITHOUT a flagship card at 2560.

Pretty sure he was talking about resolutions under 2560x. I misread it too the first time, so don't feel too bad.

[jam2k]

IMHO What Kyle is saying is, what everyone suspects, gk104 won't stand a chance against Tahiti in 2560x1600 gaming, let alone multi-monitor resolutions